search

AtomAnu / COMP0036_EPL_Prediction

0 stars 0 forks source link

Create different models for each data group/column and select the best one #5

Open AtomAnu opened 4 years ago

AtomAnu commented 4 years ago

As we have discussed, for each team, we should create different models that fit each data group/column against the number of goals scored by that team. Some data groups might be suited for classification models while some might be more suited for regression models. From these models, we can use them to make predictions and obtain the performance of each prediction. There are multiple performance measures. For classification models, the performance measures are the accuracy and the F1 Score. For regression models, the performance measures are the Root Mean Squared Error and the Mean Absolute Error. More information regarding the calculation of the performance measures can be found on pages 40-41 of this paper.

AtomAnu commented 4 years ago

Data features grouping is very essential to generate effective models. In my first attempt to group the data features given from the assignment, I have listed all the groups down below.

  1. HTHG
  2. HTAG
  3. Referee (might require more features)
  4. HS and HST (might require more features)
  5. AS and AST (might require more features)
  6. HC
  7. AC
  8. HF, HY and HR
  9. AF, AY and AR

If any of you has another idea, please don't hesitate to comment down below. Also, as a brief task allocation, here is the list of the data groups that each of us could take a look at:

@AtomAnu : 4, 8, 9 @gbifvyihk : 3, 5 @sherrydodo : 1, 2 @shuzhiYYang : 6, 7

sherrydodo commented 4 years ago

Should we add neural network as one of the models in and compare the accuracy?